Try doveadm replicator remove dummy-c-1@univ-nantes.fr
Aki
On 08/11/2021 11:32 Arnaud Abélard arnaud.abelard@univ-nantes.fr wrote:
Hello Aki,
Thanks for the tip. I changed postfix configuration so it will not use the domain part anymore and added "auth_username_format = %Ln" just in case (and reloaded dovecot). So I shouldn't have any more new user@domain users but for all the existing users, doveadm replicator status still show user@domain clone and on the new server mails are still counted twice.
Our userdb and passdb don't have the @domain defined anywhere so I don't think it's coming from there.
Is there a local cache that still has a list of those user@domain usernames that I could reset or something?
Thanks,
Arnaud
On 08/11/2021 09:48, Aki Tuomi wrote:
This sounds like you are not normalizing usernames properly.
Either use
auth_username_format = %Ln
as global setting, or return
user
attribute in both userdb and passdb lookups.Aki
On 07/11/2021 20:31 Arnaud Abélard arnaud.abelard@univ-nantes.fr wrote:
Hello again,
I've found out that some mailboxes are actually duplicated. Doveadm replicator status on the production server gives this:
~# doveadm replicator status 'dummy-c-1*' username priority fast sync full sync success sync failed dummy-c-1 none 01:13:19 01:13:19 01:13:19 - dummy-c-1@univ-nantes.fr none 00:15:28 00:15:28 00:15:28 -
That'd explain why mails are counted twice when replicated on the new server but where does this come from since I don't have this quota problem on the production server?
From the logs, it seems that postfix uses username@univ-nantes.fr when calling dovecot's LDA so I suppose that the non-'@univ-nantes.fr' mailboxes are the erroneous ones.
The users aren't duplicated in our ldap user backend and aren't using the @univ-nantes.fr part.
~# doveadm user 'dummy-c-1*' dummy-c-1
Any ideas?
Thanks,
Arnaud
On 05/11/2021 16:21, Arnaud Abélard wrote:
Hello,
We are very long time happy dovecot users (since 2008 at least). We have around 90k mailboxes and since we had to move away from our NAS storage to a ceph storage I jumped on the opportunity to enable compression with the zlib plugin and dovecot's replication mecanism. We are using debian's dovecot 2.2.27 packages on production and our new server is running dovecot's own ce-2.3.17 packages.
On the production server everything works fine but on the new server, replicated mailboxes' quota is all wrong:
on production: # doveadm quota get -u dummy-c-1 Quota name Type Value Limit % Quota Utilisateur STORAGE 660026 976563 67 Quota Utilisateur MESSAGE 8651 - 0
on new server: doveadm quota get -u dummy-c-1 Quota name Type Value Limit % Quota Utilisateur STORAGE 1125251 976563 115 Quota Utilisateur MESSAGE 16646 - 0
If I add all the S flag from the filenames n both servers I get exactly the same usage, which is coherent with the quota on the production server:
# find . -type f | grep 'S=' | awk -F'S=' '{print $2}' | awk -F',' '{print $1}' | awk -F':' '{print $1}' | paste -sd+ | bc -l 675865938
And I have exactly the same amountof mails on the two server, the replication works as expected, no unwanted duplication of mails occurs.
Of course, I've tried to ask dovecot to recalculate quotas with doveadm quota recalc -u <username>, but it doesn't fix the problem.
What am I missing?
Thanks,
Arnaud
PS: Here is my doveconf -n output:
# 2.3.17 (e2aa53df5b): /etc/dovecot/dovecot.conf # Pigeonhole version 0.5.17 (054dddfa) # OS: Linux 4.19.0-18-amd64 x86_64 Debian 10.11 # Hostname: gromel-test auth_cache_size = 10 k auth_verbose = yes disable_plaintext_auth = no doveadm_password = # hidden, use -P to show it hostname = gromel1.univ-nantes.prive lda_mailbox_autosubscribe = yes listen = * log_timestamp = "%Y-%m-%d %H:%M:%S " login_trusted_networks = (...) mail_gid = 5000 mail_location = maildir:%h mail_plugins = quota zlib notify replication mail_privileged_group = vmail mail_uid = 5000 maildir_stat_dirs = yes maildir_very_dirty_syncs = yes managesieve_notify_capability = mailto managesieve_sieve_capability = fileinto reject envelope encoded-character subaddress comparator-i;ascii-numeric relational regex imap4flags copy include variables body enotify environment mailbox date index ihave duplicate mime foreverypart extracttext namespace { inbox = yes location = prefix = INBOX. separator = . type = private } passdb { args = /etc/dovecot/dovecot-ldap.conf driver = ldap } plugin { quota = maildir:Quota Utilisateur quota_exceeded_message = Cet utilisateur a dépassé son quota, votre message n'a pu lui être livré. quota_full_tempfail = yes quota_rule = *:storage=1000M quota_rule2 = INBOX.Trash:storage=+100M sieve = ~/dovecot.sieve sieve_dir = ~/sieve sieve_extensions = -vacation sieve_global_dir = /var/lib/dovecot/sieve/global/ sieve_max_redirects = 1 zlib_save = gz zlib_save_level = 6 } postmaster_address = postmaster@<snip> protocols = imap pop3 sieve replication_max_conns = 50 service auth { client_limit = 49452 unix_listener auth-userdb { group = vmail mode = 0600 user = vmail } user = root } service dict { unix_listener dict { mode = 0600 user = vmail } } service doveadm { inet_listener { port = 12345 } } service imap-login { process_min_avail = 8 service_count = 0 user = mail } service imap { executable = imap process_limit = 16384 vsz_limit = 2 G } service managesieve-login { inet_listener sieve { port = 4190 } process_min_avail = 8 service_count = 0 user = mail vsz_limit = 2 G } service managesieve { drop_priv_before_exec = yes process_limit = 16384 } service pop3-login { process_min_avail = 8 service_count = 0 user = mail } service pop3 { drop_priv_before_exec = yes process_limit = 16384 } ssl = no userdb { args = /etc/dovecot/dovecot-ldap.conf driver = ldap } verbose_proctitle = yes protocol imap { mail_max_userip_connections = 50 mail_plugins = quota zlib notify replication imap_quota zlib } protocol pop3 { mail_plugins = quota zlib notify replication } protocol sieve { mail_max_userip_connections = 10 } protocol lda { mail_plugins = quota zlib notify replication sieve zlib }
-- Arnaud Abélard Responsable pôle Système et Stockage Service Infrastructures DSIN Université de Nantes
-- Arnaud Abélard Responsable pôle Système et Stockage Service Infrastructures DSIN Université de Nantes